094-DOLPHIN: Using RNA splicing to enhance ScRNA analysis

This paper introduces DOLPHIN, a novel deep learning framework designed to enhance single-cell RNA sequencing (scRNA-seq) analysis by moving beyond traditional gene-level quantification. The researchers explain that conventional methods often overlook crucial exon-level information and junction reads, which limits the accurate representation of cellular states. DOLPHIN integrates these detailed genomic elements by modeling genes as graph structures and processing them with a variational graph autoencoder to create superior cell embeddings. The text details how this approach significantly improves cell clustering, biomarker discovery, and alternative splicing detection, even in sparse datasets like those from 10X Genomics platforms. Ultimately, DOLPHIN aims to provide a more precise understanding of cellular heterogeneity and disease mechanisms by uncovering subtle transcriptomic differences missed by previous techniques.

References:

• Song K, Zheng Y, Zhao B, et al. DOLPHIN advances single-cell transcriptomics beyond gene level by leveraging exon and junction reads[J]. Nature Communications, 2025, 16(1): 6202.